Upgrade the CPU in the Armada 1750 laptop to a Pentium III
Information on upgrading and overclocking the Compaq Armada 1750 laptop computer. Upgrading the CPU from a Pentium II to a Pentium III 650MHz (MMC-2 socket) or faster. Some very useful information was gleaned from a Wim's Bios forum thread I stumbled across while looking for information on "MMC-2 overclocking", and then applied to upgrading and overclocking the Armada 1750.
Upgrading to a Pentium III
These are the condensed results of my research and testing, as they apply to the Compaq Armada 1750 notebook computer. I have done the upgrade myself, as detailed below, as well as have other visitors such as ACE 256 (see credits), and the results seem to be fairly consistent. If you need help in disassembling the laptop, I strongly recommend looking at my Armada 1750 Disassembly & Repair page.
The Armada 1750 laptop uses an Intel 440BX chipset, which is entirely capable of running at 100MHz front-side bus (FSB). The hardwired memory on the motherboard, at least on mine (model 6366/T/6400/D/M/1, but I think this applies to all of them), uses Toshiba TC59S64 (64Mbit) "-80" SDRAM chips, which are 8ns, and therefore PC100 capable. I found the information on interpreting the markings on the chips here. This actually makes the Armada 1750 a better candidate for upgrading and overclocking than the IBM Thinkpad 600E (which apparently uses PC66, not 100) being discussed in the forum thread I found the information on. The Armada 1750 will take a 100MHz FSB processor with an MMC-2 socket, such as the Pentium III 650 MMC2, and automatically start up at a 100MHz FSB.
Please note that the Armada 1700 uses an MMC-1 style socket, and hence is not upgradable - at least not to a Pentium III.
You will need to use CPUMSR to re-enable the L2 cache each boot (see the readme for installation instructions). I noticed the CPUMSR site was kind of slow for downloading; you can download CPUMSR v0.90 directly from Wireball.com,(required LLADrv included) if you want. CPUMSR apparently works better than the Powerleap utility I previously linked to. I have tried CPUMSR under Windows 2000 on my PIII-upgraded Armada 1750, and it works fine.
If you want CPUMSR to automatically apply the settings and exit when you start Windows, please see the documentation.
Be aware that the CPU will run at a lower clock speed on computers that do not recognize Intel Speedstep technology, which includes the Compaq Armada 1750. For example, a PIII-650MHz MMC-2 will run at 500MHz/100MHz clock/FSB. This is apparently due to the processor being multiplier locked at two different multipliers, and on a machine that does not recognize Speedstep, it will default to the lower multiplier. There is a mod to force a Speedstep processor to start with the higher multiplier, but it is reportedly a bit tricky and may not always work (it also requires physical modification of the CPU, so I don't plan on detailing it here unless I try it myself).
According to this table on Tomshardware.com, 600 and 650MHz Mobile Pentium III's will run at 500MHz ("Battery-Mode"), the PIII 700 will run at 550MHz, the 750 will run at 600MHz, 800 at 650MHz, at the 850 will run at 700MHz. The corresponding CPU "Battery Mode" voltage drop lowers the maximum thermal dissipation of the CPU close to that of what the original PII CPU dissipated. Given that the Compaq Armada 1750 is designed for a specific Thermal Dissipation maximum (TDPmax/W) of about 13.8W, judging from the PII-366, it may be a good thing from a heat standpoint (and of course a battery standpoint) that the laptop defaults to the lower speed.
Even a PIII-650 running at 500MHz is a pretty significant upgrade for this laptop, given the addition of SSE support, 50% higher front-side bus (100MHz vs. 66MHz) and therefore higher memory bandwidth, and of course the higher clock speed (and as much as 700MHz "Battery Mode" default on this machine if you can find a PIII-850 MMC-2 processor). The PIII-650 MMC2 also appears to be readily available from eBay.com for $40 or less, at the time of this writing.
You may be able to save some money by buying a 500MHz PIII MMC-2. What I gather from my research is that the 600 is the minimum speed grade that uses Speedstep. The 500 will always run at 500, from what I can tell. Availability of 500MHz Pentium III MMC-2 CPUs appears to be better, too.
In practice - performing the upgrade
I recently acquired a P3-650MHz MMC-2 CPU on eBay and upgraded my Compaq Armada 1750. Except for putting the fan in backwards the first time I reassembled the laptop (which caused the whole laptop to become noticeably warm), I had little difficulty (but I am familiar with disassembling and reassembling the laptop). The heatsink is bolted to the underside of the mainboard by three screws on the top. After removing the three screws, I gently worked the CPU free by inserting a slotted screwdriver blade between the mounting posts and the board and wiggling/prying it free a bit at a time on each post. The heatsink is bolted to the CPU by two small silver screws in among the fins. Undoing these, the heatsink comes off fairly easy (one might want to use a gentle back and forth twisting motion - I tore the thermal pad the first time I took the heatsink off).
Warning: there is a small gap between the heatsink and the CPU heatspreader - you need the mesh-reinforced thermal pad to fill this up (regular thermal grease will not make contact, and the CPU will overheat). I then transferred the mesh-reinforced thermal pad to the new Pentium III cpu, and carefully bolted on the heatsink, checking alignment and then evenly tightening the screws (well, I didn't snug them down until both were started in their holes). I then plugged the new P3 into the motherboard, and secured the screws. I noticed that the rubber (?) insulator between the small heatsink on the mainboard (covering the ATI Rage Pro graphics chip, I believe) conflicted slightly with the different component layout on the P3 CPU, but it was still possible to screw it down fully. However, this caused the board to bend a little, so I backed it off a little on the nearest support post and left a tiny gap (no more than 0.5mm). The other support posts were snugged down all the way (be careful not to overtighten). Then the laptop was reassembled.
Bootup was uneventful, aside from slightly different behavior on POST, which may have been caused by the fact I was running it on battery power without realizing it at first (the laptop powered off rather than rebooting when pressing ctrl-alt-del in the diagnostics screen), and the system information screen seemed to not want to come up (the CMOS setup still worked fine). On-the-fly speed adjustment between battery mode (run at 50% speed) and AC power never seems to have worked with my laptop and Windows 2000, and the behavior is unchanged with the new CPU - a reboot is required for the speed change to take effect. For example, after booting on battery power, and then plugging in the AC line cord, although CPU-Z reported the CPU to be running at 500MHz instead of 250MHz, performance tests still indicated the same performance as at 250MHz. After rebooting, performance was much higher.
Speaking of performance, I enabled the L2 cache using CPUMSR, (after installing the included LLADrv, required, and rebooting) and yes, CPU-Z does accurately report whether the L2 is enabled or disabled. The L2 cache is required for decent performance, and must be enabled after every startup (because it defaults to off). However, it should be possible to run CPUMSR on startup with command line commands that will automatically enable the L2 cache and quit every time you log into your computer. Check the help files included with the program, or the CPUMSR author's website.
To benchmark the new CPU (and mainly make sure that I was getting the proper level of performance from it - I could already tell it was much faster), I ran SiSoft Sandra Lite - particularly the Cache & Memory Benchmark. The PIII-500 is shown as the blue line in the graph. It's compared against CPUs with external L2 caches running at partial speed, but there is some relation to the old PII-366. As you can see, the PIII-500 starts out at over 3.5GB/sec for the L1 cache, drops off to around 2GB/sec when it hits the L2 cache, and then trails off to average of about 400MB/sec for the main memory. Not shown here (I'll add a screenshot later), but the P3 at 500MHz scores over five times higher on one of the floating point tests than the original PII-366 CPU, probably largely due to the PIII supporting SSE. Of course, this is a synthetic benchmark, so the performance increase may not be quite as drastic in the real world.
The main memory is now running at 100MHz, 2-2-2 timings. According to CPU-Z, the SPD information says the memory is intended to be running at CAS3 latency at 100MHz. However, I ran an hour of MemTest86, and detected no errors, so it appears to be stable at CAS2, 100MHz. This is a 50% increase in the frontside bus over the speed with the old PII CPU (which was a 66MHz FSB).
Unfortunately, I was only able to find information on overclocking the PIII - the original Pentium II MMC-2 CPU uses a slightly different circuit board layout and design, with differently-labelled resistor packages (instead of individual resistors), and I think that the resistor pin I want is under the heatspreader, which I don't particularly feel like drilling out the rivets on. Still, I suspect if you could figure out the equivalent resistors and wire it up, you might be able to get the 8% speed boost. I haven't tried it, though. You can cut the MMC2 socket 66/100 SEL line as detailed on page 4 of the aforementioned Wim's Bios forum thread, but it appears that few were successful in getting the PII to POST and run stably at 100MHz FSB (a 50% overclock).
The Pentium III MMC-2 CPU may be overclocked by 8% by soldering a 1Kohm resistor to the east end of R164 and the south end of capacitor C21 (when the CPU is viewed with the socket to the left, with the R153 text horizontal to your viewpoint). Resistor R153 is one of a group of three resistors, located a bit north-west from the R153 text and immediately adjacent and at right angles to R119. Note: if you solder to R157 instead, you will get a 4% overclock, if you want a lower speed for some reason. Here's a link to a photo.
The effect of performing this mod is to increase the CPUs FSB (and therefore core clock) by 8% (or 4% if you so choose). For example, a PIII-650 (defaults to 500MHz on a non-Speedstep-supporting machine) would overclock to 540MHz/108FSB. A PIII-750 (600MHz non-Speedstep / "Battery Mode") will run at 640MHz/108FSB. A PIII-850 (700MHz) would overclock to 756MHz/108FSB. In practice, this will make the machine almost* 8% faster, (for any CPU-intensive tasks) because the CPU, memory, and front-side bus are all running 8% faster. Beware! Your ram may not like running at CAS2 latency at 108MHz, and may encounter glitches. I recommend testing it with MemTest86 before and after the mod, to ensure stability.
* - (there is some overhead, so it doesn't scale exactly 1:1, but it's close)
BXtweaking :: 29 Jul '05
You might want to add something about the 440bx tweaker - it made a noticeable speed boost for me.
Note that the 440bx tweak site that ACE 256 is linking to is the Internet Archive copy of the article; it seems to have recently disappeared from sysopt.com. All the pictures and such are intact; however the agp_set program must be obtained separately from http://www.slunecnice.cz/product/AGP-INFO/. This works under 9x-type operating systems. However, it will not work under NT-type OSes (incl. 2000/XP). Furthermore, read the readme.txt file and the BX Chipset Tweaking article, and know what you are doing before you make any changes with agp_set.
Kudos go to Mike Rudkin (who has a website on upgrading and overclocking the Compaq Presario 1600 and 1200-series notebook computers) and ACE 256 -- without them this article probably never would have materialized. Mike Rudkin for e-mailing me with a query about the possibility of upgrading the CPU, and then informing me that it used an MMC-2 socket, and was not hard-wired as I had assumed. ACE 256 for e-mailing me just a day or two after Mike (and completely independently) letting me know that he had successfully upgraded the PII-300 CPU on his Compaq Armada 1750 to a Pentium III 650, and that it was running fine at 500MHz clock, 100MHz FSB, not to mention some other useful technical information (such as superior video card drivers for the laptop under 98/ME/2000/XP). Thanks guys!
All production models of the Armada 1750, so far as I know, were restricted to Pentium II CPUs with 66MHz front-side buses, so it's exciting to know that it will take a 100MHz FSB CPU.
Please note that the Armada 1700 uses an MMC-1 style socket, and hence is not upgradable - at least not to a Pentium III.
Although I attempt to make sure the information I post on this site is accurate, I make no guarantees as to success. In NO way am I responsible for any damages that may result from the useage or misuseage of the data on this, or any other page. As with most information on the Internet, you may safely assume that burning down your house is your responsibility :)
For the time being, you may post a question using the comment form below. Your question is far more likely to be answered if you include a valid e-mail address. Your e-mail address will remain private.
Previous page: Compaq Armada 1750 Laptop Upgrade, Repair & Disassembly
Next page: Armada 1750 disassembly & repair photo gallery